watr 
Light & fast WAT compiler
- feature & spec-complete, zero deps
- compile · polyfill · optimize · parse · prettify · minify
- instant wasm, JS interop, CLI, clear errors
Usage
import watr, { compile, parse, print } from 'watr'
// compile to binary
const src = '(func (export "f") (result f64) (f64.const 1))'
const binary = compile(src)
const module = new WebAssembly.Module(binary)
const { f } = new WebAssembly.Instance(module).exports
// optional, heavy transforms ship as separate entries — compose them
import optimize from 'watr/optimize' // fold constants, treeshake, eliminate dead code …
import polyfill from 'watr/polyfill' // newer features → MVP
compile(optimize(polyfill(src)))
// parse
parse('(i32.const 42)') // ['i32.const', 42]
// print
print('(module(func(result i32)i32.const 42))') // (module\n (func (result i32)\n ...
// instant wasm function
const { add } = watr`(func (export "add") (param i32 i32) (result i32)
(i32.add (local.get 0) (local.get 1))
)`
add(2, 3) // 5
// instant wasm: interpolate, auto-import ...
const { test } = watr`(func (export "test") (call ${console.log} (i32.const 42)))`
test() // logs 42
CLI
npx watr input.wat # → input.wasm
npx watr input.wat -o out.wasm # custom output
npx watr input.wat --print # pretty-print
npx watr input.wat --minify # minify
npx watr input.wat --optimize # fold, treeshake, inline, coalesce, …
npx watr input.wat --polyfill # newer features → MVP
Metrics
- watr — ~44 KB minified (~14 KB gzipped), 3,123 op/s
- spec/wast.js — 216 KB, 1,509 op/s
- wabt — 282 KB, 859 op/s
- binaryen — 1,100 KB, 473 op/s
- wat-compiler — ~152 KB (+ wabt dep), 355 op/s
Optimizer vs binaryen (wasm-opt 128)
| size (total) | time (batch) | footprint | |
|---|---|---|---|
| watr/optimize | 19,734 B | ~110 ms in-process | 149 KB min (46 KB gz) |
wasm-opt -Oz |
19,852 B | 990 ms CLI | ~1.1 MB js / native binary |
wasm-opt -O3 |
22,302 B | — | — |
Measured on test/example (21 modules, optimize(src) defaults vs wasm-opt -all).
Smaller than -O3 on every module and than -Oz in aggregate (19,734 vs 19,852 B); ties or beats -Oz on 18 of 20 files.
On a 5.5 MB module watr’s output is 1.7% SMALLER than -Oz (261,345 vs 265,977 B — cross-function outlining);
binaryen’s native core is faster there (0.44 s vs ~2.6 s optimize + 0.19 s compile) — watr’s edge is output size, batch/in-process use
with no process spawn, and a ~25× smaller footprint.
Used by
- jz – minimal static JS subset